- Home
- Skills
- Workflow Automation
- Spec Driven Development Pro
Works with the AI Agents you already use
Spec Driven Development Pro
Enforce disciplined spec-before-code workflows with living technical contracts and gated verification.
$24
Spec Driven Development Pro
Example session with this skill installed
Run the spec phase for task export-users-csv. Admins download the user list as CSV from the admin page. Columns: id, email, created_at. Max 50k rows. No deactivated users.
- Read your context and instructions
- Compiled the spec driven development
I ran the four spec phases for export-users-csv and stopped at the boundary so you can review before any code is written. Everything is in specs/export-users-csv/.
Phase results
| Phase | Output | Verdict |
|---|---|---|
| 1. Proposal | proposal.md, spec.md | PASS |
| 2. Exploration | exploration.md | PASS |
| 3. Design | design.md, tasks.md | PASS |
| 4. Behavior | spec.md refined, testing.md | PASS |
Key findings
Reuse before create (from exploration.md):
| Need | Exists | Path | Decision |
|---|---|---|---|
| File download helper | yes | src/shared/download/saveBlob.ts | reuse |
| CSV serializer | no | - | create in src/shared/csv/ |
| Active users query | no | - | create in src/entities/user/ |
Requirements (from spec.md):
- The system SHALL export id, email and created_at for every active user.
- The system MUST exclude deactivated users.
- The system MUST reject exports above 50,000 rows with a clear error.
Tasks (from tasks.md, lowest layer first):
- T1. Add CSV serializer with quote escaping (
shared) - T2. Add active users query (
entities) - T3. Add export flow with the 50,000 row limit (
features) - T4. Wire the admin button (
pages) - T5. Cover all 7 scenarios from
testing.md
Next steps
- Review the four markdown files. Specs are the contract: implementation will not edit them.
- If you change your mind, use the amend flow before any code is touched.
- When approved, run the build phase: implementation, strict review, then the commit plan.
PASS
Connects securely to your tools. The creator never sees your data.
About this skill
Development teams often struggle with feature creep and architectural drift where code quickly diverges from original requirements. This skill enforces a disciplined, two-phase software development lifecycle that treats specifications as a living, immutable contract during implementation.
What it does
- Enforces Spec-Driven Development by separating the planning phase (proposal, exploration, design) from the build phase.
- Maintains living contracts by treating
spec.mdanddesign.mdas read-only during the coding phase to prevent silent drift. - Generates structured artifacts including a task-specific folder containing proposals, test matrices, task lists, and commit plans.
- Ensures zero guesswork by requiring the agent to verify file existence and content before making any technical assumptions.
- Provides strict verification through a mandatory verdict system (
PASS,GAP, orVETO) at the end of every phase.
How it works
- Initiate the task by creating a task slug and defining the objective in
task.md. - Execute the Spec phase by running through proposal, exploration, design, and behavior specification steps.
- Execute the Build phase to implement code, perform a strict review against the specs, and generate a commit plan.
- Handle changes via a formal amend flow that surgically updates specs and logs changes before implementation resumes.
Frameworks & tools
This skill is framework-agnostic and works with any codebase. It relies on Markdown for documentation and standard CLI tools for linting, typechecking, and testing verification.
Why this beats prompting it yourself
Standard prompts often let the LLM rush into coding, leading to architectural debt and missing edge cases. This skill enforces a multi-step gated process that requires explicit evidence and verification before a single line of application code is written.
Use cases
- Non-trivial features where the implementation details need to be agreed upon before starting work.
- Complex refactors that require a clear exploration of dependencies and side effects.
- Strict code reviews that need to be audited against an original technical design rather than just syntax rules.
Known limitations
Does not perform history-mutating Git commands like commit or push. Requires the user to manually review specs between the planning and build phases.
No history-mutating Git commands (add, commit, push). User must review artifacts between spec and build phases. Zero guesswork policy means it will stop if info is missing.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 19 days ago
Requires